PsyMuKB: An Integrative De Novo Variant Knowledge Base for Developmental Disorders

Inferring the Effects of Protein Variants on Protein-Protein Interactions with Interpretable Transformer Representations

Identifying pathogenetic variants and inferring their impact on protein-protein interactions sheds light on their functional consequences on diseases. Limited by the availability of experimental data on the consequences of protein interaction, most existing methods focus on building models to predict changes in protein binding affinity. Here, we introduced MIPPI, an end-to-end, interpretable transformer-based deep learning model that learns features directly from sequences by leveraging the interaction data from IMEx. MIPPI was specifically trained to determine the types of variant impact (increasing, decreasing, disrupting, and no effect) on protein-protein interactions. We demonstrate the accuracy of MIPPI and provide interpretation through the analysis of learned attention weights, which exhibit correlations with the amino acids interacting with the variant. Moreover, we showed the practicality of MIPPI in prioritizing de novo mutations associated with complex neurodevelopmental disorders and the potential to determine the pathogenic and driving mutations. Finally, we experimentally validated the functional impact of several variants identified in patients with such disorders. Overall, MIPPI emerges as a versatile, robust, and interpretable model, capable of effectively predicting mutation impacts on protein-protein interactions and facilitating the discovery of clinically actionable variants.

AutoCaSc: Prioritizing candidate genes for neurodevelopmental disorders

Routine exome sequencing (ES) in individuals with neurodevelopmental disorders (NDD) remains inconclusive in >50% of the cases. Research analysis of unsolved cases can identify novel candidate genes but is time-consuming, subjective, and hard to compare between labs. The field, therefore, requires automated and standardized assessment methods to prioritize candidates for matchmaking. We developed AutoCaSc (https://autocasc.uni-leipzig.de) based on our candidate scoring scheme. We validated our approach using synthetic trios and real in-house trio ES data. AutoCaSc consistently (94.5% of all cases) scored the relevant variants in valid novel NDD genes in the top three ranks. In 93 real trio exomes, AutoCaSc identified most (97.5%) previously manually scored variants while evaluating additional high-scoring variants missed in manual evaluation. It identified candidate variants in previously undescribed NDD candidate genes (CNTN2, DLGAP1, SMURF1, NRXN3, and PRICKLE1). AutoCaSc enables anybody to quickly screen a variant for its plausibility in NDD. After contributing >40 descriptions of NDD-associated genes, we provide usage recommendations based on our extensive experience. Our implementation is capable of pipeline integration and therefore allows the screening of large cohorts for candidate genes. AutoCaSc empowers even small labs to a standardized matchmaking collaboration and to contribute to the ongoing identification of novel NDD entities.

Embryonic Deletion of TXNIP in GABAergic Neurons Enhanced Oxidative Stress in PV+ Interneurons in Primary Somatosensory Cortex of Aging Mice: Relevance to Schizophrenia

The brain is susceptible to perturbations of redox balance, affecting neurogenesis and increasing the risks of psychiatric disorders. Thioredoxin-interacting protein (TXNIP) is an endogenous inhibitor of the thioredoxin antioxidant system. Its deletion or inhibition suggests protection for a brain with ischemic stroke or Alzheimer’s disease. Combined with conditional knockout mice and schizophrenia samples, we aimed to investigate the function of TXNIP in healthy brain and psychiatric disorders, which are under-studied. We found TXNIP was remarkedly expressed in the prefrontal cortex (PFC) during healthy mice’s prenatal and early postnatal periods, whereas it rapidly decreased throughout adulthood. During early life, TXNIP was primarily distributed in inhibitory and excitatory neurons. Contrary to the protective effect, the embryonic deletion of TXNIP in GABAergic (gamma-aminobutyric acid-ergic) neurons enhanced oxidative stress in PV⁺ interneurons of aging mice. The deleterious impact was brain region-specific. We also investigated the relationship between TXNIP and schizophrenia. TXNIP was significantly increased in the PFC of schizophrenia-like mice after MK801 administration, followed by oxidative stress. First episode and drug naïve schizophrenia patients with a higher level of plasma TXNIP displayed severer psychiatric symptoms than patients with a low level. We indicated a bidirectional function of TXNIP in the brain, whose high expression in the early stage is protective for development but might be harmful in a later period, associated with mental disorders.

De novo mutations identified by whole-genome sequencing implicate chromatin modifications in obsessive-compulsive disorder

Obsessive-compulsive disorder (OCD) is a chronic anxiety disorder with a substantial genetic basis and a broadly undiscovered etiology. Recent studies of de novo mutation (DNM) exome-sequencing studies for OCD have reinforced the hypothesis that rare variation contributes to the risk. We performed, to our knowledge, the first whole-genome sequencing on 53 parent-offspring families with offspring affected with OCD to investigate all rare de novo variants and insertions/deletions. We observed higher mutation rates in promoter-anchored chromatin loops (empirical P = 0.0015) and regions with high frequencies of histone marks (empirical P = 0.0001). Mutations affecting coding regions were significantly enriched within coexpression modules of genes involved in chromatin modification during human brain development. Four genes—SETD5, KDM3B, ASXL3, and FBL—had strong aggregated evidence and functionally converged on transcription’s epigenetic regulation, suggesting an important OCD risk mechanism. Our data characterized different genome-wide DNMs and highlighted the contribution of chromatin modification in the etiology of OCD.

Dissection of the Genetic Association between Anorexia Nervosa and Obsessive-Compulsive Disorder at the Network and Cellular Levels

Anorexia nervosa (AN) and obsessive-compulsive disorder (OCD) exhibit a high co-morbidity rate, similar symptoms, and a shared genetic basis. However, an understanding of the specific underlying mechanisms of these commonalities is currently limited. Here, we collected Genome-Wide Association Analysis results for AN and OCD, and obtained genes hit by the top SNPs as the risk genes. We then carried out an integrative coexpression network analysis to explore the convergence and divergence of AN and OCD risk genes. At first, we observed that the AN risk genes were enriched in coexpression modules that involved extracellular matrix functions and highly are expressed in the postnatal brain, limbic system, and non-neuronal cell types, while the OCD risk genes were enriched in modules of synapse function, the prenatal brain, cortex layers, and neurons. Next, by comparing the expressions from the eating disorder and OCD postmortem patient brain tissues, we observed both disorders have similar prefrontal cortex expression alterations influencing the synapse transmission, suggesting that the two diseases could have similar functional pathways. We found that the AN and OCD risk genes had distinct functional and spatiotemporal enrichment patterns but carried similar expression alterations as a disease mechanism, which may be one of the key reasons they had similar but not identical clinical phenotypes.

Alternatively Splicing Interactomes Identify Novel Isoform-Specific Partners for NSD2

Nuclear receptor SET domain protein (NSD2) plays a fundamental role in the pathogenesis of Wolf-Hirschhorn Syndrome (WHS) and is overexpressed in multiple human myelomas, but its protein-protein interaction (PPI) patterns, particularly at the isoform/exon levels, are poorly understood. We explored the subcellular localizations of four representative NSD2 transcripts with immunofluorescence microscopy. Next, we used label-free quantification to perform immunoprecipitation mass spectrometry (IP-MS) analyses of the transcripts. Using the interaction partners for each transcript detected in the IP-MS results, we identified 890 isoform-specific PPI partners (83% are novel). These PPI networks were further divided into four categories of the exon-specific interactome. In these exon-specific PPI partners, two genes, RPL10 and HSPA8, were successfully confirmed by co-immunoprecipitation and Western blotting. RPL10 primarily interacted with Isoforms 1, 3, and 5, and HSPA8 interacted with all four isoforms, respectively. Using our extended NSD2 protein interactions, we constructed an isoform-level PPI landscape for NSD2 to serve as reference interactome data for NSD2 spliceosome-level studies. Furthermore, the RNA splicing processes supported by these isoform partners shed light on the diverse roles NSD2 plays in WHS and myeloma development. We also validated the interactions using Western blotting, RPL10, and the three NSD2 (Isoform 1, 3, and 5). Our results expand gene-level NSD2 PPI networks and provide a basis for the treatment of NSD2-related developmental diseases.

Integrative Analysis Identified Key Schizophrenia Risk Factors from an Abnormal Behavior Mouse Gene Set

Schizophrenia (SCZ) is a severe chronic psychiatric illness with heterogeneous symptoms. However, the pathogenesis of SCZ is unclear, and the number of well-defined SCZ risk factors is limited. We hypothesized that an abnormal behavior (AB) gene set verified by mouse model experiments can be used to better understand SCZ risks. In this work, we carried out an integrative bioinformatics analysis to study two types of risk genes that are either differentially expressed (DEGs) in the case-control study data or carry reported SCZ genetic variants (MUTs). Next, we used RNA-Seq expression data from the hippocampus (HIPPO) and dorsolateral prefrontal cortex (DLPFC) to define the key genes affected by different types (DEGs and MUTs) in different brain regions (DLPFC and HIPPO): DLPFC-kDEG, DLPFC-kMUT, HIPPO-kDEG, and HIPPO-kMUT. The four hub genes (SHANK1, SHANK2, DLG4, and NLGN3) of the biological functionally enriched terms were strongly linked to SCZ via gene co-expression network analysis. Then, we observed that specific spatial expressions of DLPFC-kMUT and HIPPO-kMUT were convergent in the early stages and divergent in the later stages of development. In addition, all four types of key genes showed significantly larger average protein-protein interaction degrees than the background. Comparing the different cell types, the expression of four types of key genes showed specificity in different dimensions. Together, our results offer new insights into potential risk factors and help us understand the complexity and regional heterogeneity of SCZ.